Carbonation is a key mechanism for deterioration of reinforced concrete structures. In design for resistance to carbonation, the performance of mixes is often specified and measured in laboratory carbonation tests where initial carbonation is sometimes present in specimens. The coherent interpretation of results from carbonation tests is vital as they enable a more accurate prediction of the performance of concrete structures in situ. This paper assesses two different approaches to considering the initial carbonation depth when extracting the carbonation coefficient from results of carbonation testing. Experimental data is compared to models for either approach using least squares regression. Both linear and non-linear representations of the initial carbonation depth are shown to fit the data well. The non-linear approach gives a larger estimate of carbonation coefficient than the linear approach, and is more consistent with the mathematical derivation of the carbonation equation. The ramifications of this difference will be most significant when the initial carbonation depth is large relative to the depth of the carbonation front. The accurate modelling of carbonation progression underpins performance-based design of new concrete structures and the assessment of existing concrete infrastructure.

Development of robust concrete mixes with a lower environmental impact is challenging due to natural variability in constituent materials and a multitude of possible combinations of mix proportions. Making reliable property predictions with machine learning can facilitate performance-based specification of concrete, reducing material inefficiencies and improving the sustainability of concrete construction. In this work, we develop a machine learning algorithm that can utilize intermediate target variables and their associated noise to predict the final target variable. We apply the methodology to specify a concrete mix that has high resistance to carbonation, and another concrete mix that has low environmental impact. Both mixes also fulfill targets on the strength, density, and cost. The specified mixes are experimentally validated against their predictions. Our generic methodology enables the exploitation of noise in machine learning, which has a broad range of applications in structural engineering and beyond.